Himshikhar Mazumdar

Himshikhar Mazumdar is a Research Associate at the GIScience Research Group of the Institute of Geography in Heidelberg with a background in Physics specialized in Astrophysics from the University of Heidelberg. His research focuses on estimating traffic volumes using neural networks, estimating vehicle velocity using satellite imaging and using OSM for highway road network analysis, vandalism detection etc.

Research Interests

• Big Spatial Data Analysis
• Neural networks and Machine Learning

Projects

• Geco Project: Estimating traffic volumes using neural networks & machine learning
• PITS: Planetary Image Traffic Speeds: Using Planet satellite images to estimate traffic velocity on highways
• OSM road monitor use case: We are investigating the access to education across countries globally. This analysis relies on a) the OSM road network and b) the location of education facilities (e.g. primary schools) in OSM.
• OSM road monitor: Road data updates are essential for navigation, logistics, and emergency response, but official and commercial sources often have slow update cycles.
  - OpenStreetMap (OSM) provides rapid updates, as seen in disaster scenarios, but issues like vandalism, satellite data processing delays, and infrequent verification still limit its effectiveness.
  - Many routing algorithms struggle with continuously changing data, reducing efficiency and requiring less optimal methods like Dijkstra's algorithm.

Education

Master of Science, Physics (Astrophysics) 2021
University of Heidelberg – Heidelberg, Germany
Physics (Astrophysics)
Specialized in astrophysics with a focus on Python programming.
MSc. Dissertation: Adaptation of kinetic field theory to a non-expanding space-time background. Developed a C++ code for numerical analysis.
Gained expertise in data analysis, data processing, visualization, and numerical coding.

Bachelor of Science, Physics (Research), 2019
Shiv Nadar University – Greater Noida, India
Physics
BSc. Dissertation: Statistical investigation of the relationship between the clump mass function and the initial mass function in star formation (Astronomy).

Professional Occupation
05/2023 – 31/01/2025 Research Assistant, HeiGIT gGmbH, Heidelberg, Germany

• Developed and implemented machine learning models to predict average daily traffic in road networks.
• Created scripts for geospatial data processing to enhance data handling efficiency.
• Worked in projects with an agile working style and flexible team environments (floating teams).
• Participated in a deep learning production workshop covering model preparation, integration, serving techniques, and API development.
• Collaborated in a team developing a plugin for calculating carbon emissions from land use and land cover changes over a given area and observation period.

Publications

• Detection and Velocity Estimation of Moving Vehicles on Roads Using Planetscope SuperDove Imagery
  Adamaik Maciej, Grinblat Yulia, Julian Psotta, Nir Fulman, Mazumdar Himshikhar, Shiyu Tang and Alexander Zipf